c00cjz00/ft_dataset_parquet

• 89.7k筆 b8.3-patch3: b8.3-patch3.parquet
• 89.7k筆 b8.3-patch3: b8.3-patch3.parquet
• 89.7k筆 b8.3-patch3: b8.3-patch3.parquet
• 21.2k筆 scienceqa_zh: scienceqa_zh_train.parquet, scienceqa_zh_test.parquet, scienceqa_zh_validation.parquet

b8.3-patch3

• Based on b8.3-patch2, replace two translation dataset.
  • Remove ultrachat-MixtralTranslation and ultrachat-GoogleTranslation
  • Add GPT4o-Translation-zh2en and GPT4o-Translation-en2zh.

    Source	Numbers
    GPT4o-Translation-zh2en	12,408
    GPT4o-Translation-en2zh	12,324

b8.3patch2

• Based on b8.3-patch1, less translation data (10k for zh2en+en2zh).
• Remove taide_taiwan_role_dialogues
• New translated csqa.

b8.3-patch1

• Based on b8.3, replace the ZH->EN translation dataset with the mixtral8*7B version.
• Data Source Numbers

  Source	Numbers
  ultrachat-tw-value	42,134
  dolphin-coder	22,520
  taide_wizardlm	19,781
  capybara	16,006
  ultrachat-MixtralTranslation	12,134
  ultrachat-GoogleTranslation	11,212
  ultrachat-CreationAndWriting	6,768
  taide_taiwan_role_dialogues	5,663
  ultrachat-Extraction	4,953
  capybara-knowlogic	1,712
  ultrachat-Summary	1,000
  csqa	738

b8.1

• b8.0 + tw_song + who_are_you

b8.0

{
  "self-identity": {
    "num_ratio": 0.01,
    "n_clusters": 20,
    "data": {
      "taide-self-identity": {
        "dataset_kwargs": {
          "path": "TLLM/ft-taide-self-identity",
          "split": "train"
        },
        "message_col": "messages",
        "strategy": "concat"
      }
    }
  },
  "multi_turn": {
    "num_ratio": 0.35,
    "n_clusters": 50,
    "data": {
      "taide_taiwan_role_dialogues": {
        "dataset_kwargs": {
          "path": "TLLM/ft-formated-data",
          "name": "taide_taiwan_role_dialogues",
          "split": "train"
        },
        "strategy": "concat",
        "message_col": "messages"
      },
      "ultrachat-tw-value": {
        "dataset_kwargs": {
          "path": "TLLM/UltraChat",
          "name": "QuestionAboutTheWorld",
          "split": "train"
        },
        "strategy": "concat",
        "message_col": "messages"
      }
    }
  },
  "office_tasks": {
    "num_ratio": 0.35,
    "n_clusters": 50,
    "data": {
      "ultrachat-CreationAndWriting": {
        "dataset_kwargs": {
          "path": "TLLM/UltraChat",
          "name": "CreationAndWriting",
          "split": "train"
        },
        "strategy": "concat",
        "message_col": "messages"
      },
      "ultrachat-AssistanceOnExistingMaterials": {
        "dataset_kwargs": {
          "path": "TLLM/UltraChat",
          "name": "AssistanceOnExistingMaterials",
          "split": "train"
        },
        "strategy": "concat",
        "message_col": "messages"
      }
    }
  },
  "single_turn": {
    "num_ratio": 0.15,
    "n_clusters": 30,
    "data": {
      "taide_wizardlm": {
        "dataset_kwargs": {
          "path": "TLLM/ft-taide-wizardLM",
          "split": "train"
        },
        "strategy": "concat",
        "message_col": "messages"
      },
      "csqa": {
        "dataset_kwargs": {
          "path": "TLLM/ft-formated-data",
          "name": "csqa-cot-zh",
          "split": "train"
        },
        "strategy": "concat",
        "message_col": "messages"
      }
    }
  },
  "coding": {
    "num_ratio": 0.14,
    "n_clusters": 40,
    "data": {
      "dolphin-coder": {
        "dataset_kwargs": {
          "path": "TLLM/ft-formated-data",
          "name": "dolphin-coder",
          "split": "train"
        },
        "strategy": "concat",
        "message_col": "messages"
      }
    }
  }
}